diff --git c/PKGBUILD i/PKGBUILD
index 9fc3617..2b46fc4 100644
--- c/PKGBUILD
+++ i/PKGBUILD
@@ -1,9 +1,9 @@
 # Maintainer: Rafael Dominiquini <rafaeldominiquini at gmail dot com>
 
 pkgbase=ollama-bin
-pkgname=(ollama-bin ollama-cuda12-bin ollama-cuda13-bin ollama-mlx-cuda13-bin ollama-vulkan-bin)
+pkgname=(ollama-bin ollama-cuda12-bin ollama-cuda13-bin ollama-vulkan-bin)
 pkgver=0.23.1
-pkgrel=1
+pkgrel=2
 pkgdesc="Create, run and share large language models (LLMs)"
 arch=('x86_64' 'aarch64')
 _barch=('amd64' 'arm64')
@@ -21,15 +21,21 @@ optdepends=("ollama-cuda12: NVIDIA GPU Support"
 source=("LICENSE-${pkgver}::${_urlraw}/LICENSE"
         "README-${pkgver}.md::${_urlraw}/README.md"
         "ollama.conf"
+        "ollama-cuda.conf"
         "ollama.service"
+        "ollama-cuda.service"
+        "ollama-vulkan.service"
         "sysusers.conf"
         "tmpfiles.d")
 source_x86_64=("ollama-${arch[0]}-${pkgver}.tzst::${url}/releases/download/v${pkgver}/ollama-linux-${_barch[0]}.tar.zst")
 source_aarch64=("ollama-${arch[1]}-${pkgver}.tzst::${url}/releases/download/v${pkgver}/ollama-linux-${_barch[1]}.tar.zst")
 sha256sums=('5934ed2ce0d15154bcdb9c85203210abac0da4314af34081e36df4599f90b226'
             'f900b4ece4a4de07e36a192ae642f29395114443f2f1928689fd8c776f1ba2f1'
-            '2503546a6d26559bce06ba6c61100026d85864b4c49bd6e4c80c596c5d22e197'
+            '964956597a56ab8a27f081510b41fb1d60e22c543bbd7267e23dca4ce3334461'
+            '5a502d9eab47078f9136b127d72289b3083934c222c7735c4585d0dbc3fef893'
             '24871ffd940212e04e9bd3c334cfd4e3c4e845b374c5d0ed369fd32496b05fdb'
+            '138f38a3ab5582d7a2f9f21c98a8a1112e3afb767e32cc348cb320938874104b'
+            'a47d96ee424ff4a113e9d60f03805dfa9ddb25df5e54fc242306b9b20b0bae0d'
             '14e2e267be85b6943f66dfe60e73f5e0a611eaf40ee69a4cc0d497d071392cf4'
             '137e1d50a5f3058c30a73b7bb3c323888d225e6a7ae47564be869827db0659a3')
 sha256sums_x86_64=('1079dad63e0e0f2d3279280ce6d8a93f3af53e79afb14f74813e0b35d9b96d54')
@@ -72,6 +78,8 @@ package_ollama-cuda12-bin() {
 
     cd "${srcdir}/" || exit
 
+    install -Dm644 "./ollama-cuda.service" "${pkgdir}/usr/lib/systemd/system/ollama.service.d/01-cuda.conf"
+    install -Dm644 "./ollama-cuda.conf" "${pkgdir}/etc/ollama-cuda.conf"
     install -dm755 "${pkgdir}/usr/lib/ollama/"
     for lib in "./lib/ollama/cuda_v12/lib"*; do
         cp -P "${lib}" "${pkgdir}/usr/lib/ollama/"
@@ -83,34 +91,18 @@ package_ollama-cuda13-bin() {
 
     provides=("ollama-cuda13")
     depends=("ollama-bin" "nvidia-libgl")
-    optdepends=("ollama-mlx-cuda13-bin: Image Generation")
     conflicts=("ollama-cuda" "ollama-cuda12" "ollama-vulkan")
 
     cd "${srcdir}/" || exit
 
+    install -Dm644 "./ollama-cuda.service" "${pkgdir}/usr/lib/systemd/system/ollama.service.d/01-cuda.conf"
+    install -Dm644 "./ollama-cuda.conf" "${pkgdir}/etc/ollama-cuda.conf"
     install -dm755 "${pkgdir}/usr/lib/ollama/"
     for lib in "./lib/ollama/cuda_v13/lib"*; do
         cp -P "${lib}" "${pkgdir}/usr/lib/ollama/"
     done
 }
 
-package_ollama-mlx-cuda13-bin() {
-    pkgdesc='Create, run and share large language models (LLMs) with MLX CUDA 13 (Image Generation)'
-
-    provides=("ollama-mlx")
-    depends+=("ollama-cuda13")
-    conflicts=("ollama-cuda" "ollama-cuda12" "ollama-vulkan")
-
-    cd "${srcdir}/" || exit
-
-    install -dm755 "${pkgdir}/usr/lib/ollama/"
-    for lib in "./lib/ollama/mlx_cuda_v13/lib"*; do
-        cp -P "${lib}" "${pkgdir}/usr/lib/ollama/"
-    done
-
-    find "${pkgdir}/usr/lib/ollama/" -xtype l -delete
-}
-
 package_ollama-vulkan-bin() {
     pkgdesc='Create, run and share large language models (LLMs) with Vulkan'
 
@@ -120,6 +112,7 @@ package_ollama-vulkan-bin() {
 
     cd "${srcdir}/" || exit
 
+    install -Dm644 "./ollama-vulkan.service" "${pkgdir}/usr/lib/systemd/system/ollama.service.d/01-vulkan.conf"
     install -dm755 "${pkgdir}/usr/lib/ollama/"
     for lib in "./lib/ollama/vulkan/lib"*; do
         cp -P "${lib}" "${pkgdir}/usr/lib/ollama/"
diff --git c/ollama-cuda.conf i/ollama-cuda.conf
new file mode 100644
index 0000000..30ae011
--- /dev/null
+++ i/ollama-cuda.conf
@@ -0,0 +1,9 @@
+# CUDA Environment Variables for Ollama Optimization
+CUDA_VISIBLE_DEVICES=0,1,2,3
+CUDA_DEVICE_ORDER=PCI_BUS_ID
+CUDA_LAUNCH_BLOCKING=0
+CUDA_MODULE_LOADING=LAZY
+CUDA_MODULE_DATA_LOADING=LAZY
+CUDA_CACHE_MAXSIZE=2147483648
+CUDA_CACHE_PATH=/var/cache/cuda
+CUDA_LOG_FILE=/var/log/cuda.log
diff --git c/ollama-cuda.service i/ollama-cuda.service
new file mode 100644
index 0000000..9e81ebe
--- /dev/null
+++ i/ollama-cuda.service
@@ -0,0 +1,2 @@
+[Service]
+EnvironmentFile=-/etc/ollama-cuda.conf
diff --git c/ollama-vulkan.service i/ollama-vulkan.service
new file mode 100644
index 0000000..3fc8359
--- /dev/null
+++ i/ollama-vulkan.service
@@ -0,0 +1,2 @@
+[Service]
+Environment="OLLAMA_VULKAN=1"
diff --git c/ollama.conf i/ollama.conf
index 2038e1f..6a1d469 100644
--- c/ollama.conf
+++ i/ollama.conf
@@ -1,13 +1,3 @@
-# CUDA Environment Variables for Ollama Optimization
-CUDA_VISIBLE_DEVICES=0,1,2,3
-CUDA_DEVICE_ORDER=PCI_BUS_ID
-CUDA_LAUNCH_BLOCKING=0
-CUDA_MODULE_LOADING=LAZY
-CUDA_MODULE_DATA_LOADING=LAZY
-CUDA_CACHE_MAXSIZE=2147483648
-CUDA_CACHE_PATH=/var/cache/cuda
-CUDA_LOG_FILE=/var/log/cuda.log
-
 # Ollama Environment Variables
 OLLAMA_HOST="http://127.0.0.1:11434"
 OLLAMA_KEEP_ALIVE="5m"